Obesity in C57BL/6J mice fed diets differing in carbohydrate and fat but not energy content

Abstract

To investigate the contributions of carbohydrate and fat to obesity we measured the body weight, body composition and food intake of adult C57BL/6J mice fed ad libitum with various combinations of two semisynthetic diets that differed in carbohydrate and fat but not in protein, micronutrient or energy content. In Experiment 1, involving male mice, body weights were similar in groups fed diets comprised of (by energy) 20% protein, 75% carbohydrate and 5% fat (C75-F5) or 20% protein, 5% carbohydrate and 75% fat (C5-F75). However, mice fed a 50:50 composite mixture of the C75-F5 and C5-F75 diets (i.e., a C40-F40 diet) became substantially more obese. Mice that could choose between the C75-F5 and C5-F75 diets ate equal amounts of each diet and gained almost as much weight as did the group fed C40-F40 diet. Mice switched every day between the C75-F5 and C5-F75 diets gained no more weight than did those fed either diet exclusively. In Experiment 2, male and female mice were fed chow or one of 8 isocaloric diets that differed parametrically in carbohydrate and fat content. Groups fed diets in the middle of the range (i.e., C35-F45 or C45-F35) weighed significantly more and were significantly fatter than were those fed diets with more extreme proportions of carbohydrate and fat (e.g., C75-F5, C5-F75), an effect that was more pronounced in males than females. In Experiment 3 and 4, male mice fed versions of the C40-F40 formulation gained more weight than did those fed the C75-F5 or C5-F75 formulations irrespective of whether the carbohydrate was predominantly sucrose or predominantly starch, or whether the fat was vegetable shortening, corn oil, palm oil or canola oil; the type of carbohydrate or fat had little or no impact on body weight. In all four experiments, energy intakes differed among the diet groups but could not account for the differences in body weight. These results demonstrate that the proportion of carbohydrate and fat in the diet influences body weight independently of energy content, and that the type of carbohydrate or fat has little impact on body weight. Consuming carbohydrate and fat simultaneously or in close temporal proximity exacerbates obesity.

1.1. INTRODUCTION

Many investigators have observed that laboratory rodents become obese if fed diets high in carbohydrate, fat, or both carbohydrate and fat [e.g., (1,4,6–10,14,16,19,23,24,27–29,34,36,37,40,46,48,49,53,57)]. These observations are often cited as proof that carbohydrate or fat cause obesity but cautious interpretation is required because usually the diets employed are high in energy as well as one or both macronutrients. Undiluted carbohydrate and fat are energy-dense so adding them to a typical rodent diet increases its energy content. This is particularly true for fat because fat has more than double the energy density of carbohydrate or protein (i.e., ~9 vs. ~4 kcal/g). Thus, macronutrient content and energy content are confounded; the obesity could arise from the added energy rather than the added carbohydrate or fat per se.

To implicate carbohydrate or fat in obesity requires dissociating the macronutrients from their energy content. This has been attempted by pair-feeding [e.g., (2,3,12,30,32,36,42)], controlled infusion of macronutrients (28), or adding a nonnutritive filler such as cellulose to maintain diets isocaloric [e.g., (56)]. Here, we used the latter method. We formulated a pair of semisynthetic diets that differed in carbohydrate, fat and cellulose content but not in protein, micronutrient or energy content. One diet was (by energy) 20% protein, 75% carbohydrate and 5% fat (abbreviated C75-F5); the other was 20% protein, 5% carbohydrate and 75% fat (abbreviated C5-F75). In Experiment 1, reported here, we found that mice fed a composite mixture of these two diets gained more body weight and fat than did mice fed either diet alone. This led us to conduct three more experiments. One was a carbohydrate-fat “dose-response” experiment. For this, we tracked the body weights of male and female mice fed one of a range of eight isocaloric diets that differed parametrically in carbohydrate and fat. The other two experiments investigated whether the type of carbohydrate (sucrose or starch) or fat (vegetable shortening, corn oil, palm oil or canola oil) was particularly effective at increasing body weight. The results suggest that peak obesity occurs with a 50:50 carbohydrate:fat mixture, and that the type of carbohydrate or fat makes little-or-no contribution to obesity with these isocaloric diets.

2.1. METHODS

Protocols of the experiments were approved by the IACUC of the Monell Chemical Senses Center. Methods followed the principles outlined in the National Research Council’s Guide for the Care and Use of Laboratory Animals, 8th edition (25).

2.1.1. Subjects and maintenance

The experiments involved C57BL/6J mice purchased from The Jackson Laboratory (Bar Harbor, ME; stock number 000664). The mice were kept in a vivarium at ^~23°C with fluorescent illumination between 0700 – 1900 h. They were housed individually in plastic tub cages measuring 26.5 cm x 17 cm x 12 cm. Each cage had a stainless steel wire lid that included a hopper to hold pelleted food and space to hold a glass water bottle with a neoprene stopper and stainless steel spout [see (52) for details]. Food (described below) and deionized water were available ad libitum. Aspen wood chips were scattered on the cage floor except when food intake measurements were being made (see below). The mice were transferred to clean cages with refilled food and fresh wood chips once every week, at the time they were weighed (see below).

2.1.2. Diets

When mice first arrived in our animal facility they were fed Teklad 8604 diet [CHOW; a cereal-based chow containing 32% protein, 14% fat and 54% carbohydrate (by energy content); 3.0 kcal/g according to the product description; Envigo (15)]. In all four experiments, a group of mice was fed CHOW throughout the 8-16-week period the other mice were fed semisynthetic diets. This CHOW-fed group provided an external control to help identify sporadic problems that might affect body weight (e.g., uncontrolled changes in vivarium temperature or humidity due to physical plant malfunctions). Comparing growth curves (i.e., body weights) of the CHOW-fed groups across experiments ensured there were no large experiment-to-experiment variations affecting food intake and/or body weight (e.g., seasonal effects). The CHOW-fed groups also provided benchmarks to compare body weight gains with the groups fed semisynthetic diets.

The semisynthetic diets we used are described in Tables 1–3. The formulations were modified from the American Institute of Nutrition 1976 amended (AIN-76A) recommendations. In every diet used here, 20% of energy was provided as protein (casein). Carbohydrate was provided as a mixture of sucrose and cornstarch. In Experiment 3, which included some diets with very little starch, all diets contained Dyetrose, which is a proprietary maltodextrin that facilitates pelleting. Fat was provided as vegetable shortening (in all experiments) or corn oil, palm oil or canola oil (in Experiment 4). Because fat is more energy-dense than carbohydrate, appropriate amounts of cellulose were added to maintain energy content constant at 3.77 kcal/g (experiment 1, 2, and 4) or 3.62 kcal/g (experiment 3). The calculations of the proportion of energy provided by each ingredient are given in Tables 1–3 and accounted for the sucrose provided as a carrier for the mineral and vitamin mixes.

Table 1.

Ingredients of the diets used in Experiment 1 and 4.

Ingredient, kcal/g	Diet name abbreviation and Dyets catalog no.
	C75-F5 104586	C40-F40 104588	C5-F75 104587
Casein, 3.58	207.5	207.5	207.5
DL-Methionine, 4.0	3.0	3.0	3.0
Cornstarch, 3.6	163.8	85.8	7.8
Sucrose, 4.0	546.1	286.1	26.1
Vegetable shorteninga, 9.0	21.0	167.7	314.5
Cellulose, 0.0	11.1	202.3	393.5
Mineral mixb, 0.47	35.0	35.0	35.0
Vitamin mixc, 3.92	10.0	10.0	10.0
Choline chloride, 0.0	2.0	2.0	2.0
Dye, 0.0 (green/blue/red)	0.5	0.5	0.5
Total energy, kcal/kg	3.77	3.77	3.77
Percent energy as:
Protein %	20	20	20
CHO %	75	40	5
Fat %	5	40	75

Notes: Values in the body of the table are grams per kilogram diet. The diets were prepared by Dyets Inc. (Bethlehem, PA) in 50-kg lots. Batches of 1-2 kg were stored in a cold room at 4 C until needed.

^aPrimex brand. In some groups in Experiment 4, vegetable shortening was replaced with corn oil, palm oil or canola oil.

^bAIN-76A mineral mix #200000 contains (g/kg mix): Calcium phosphate, dibasic (500), NaCl (74), potassium citrate■H2O (220), potassium sulfate (52), magnesium oxide (24), manganous carbonate (3.5), ferric citrate U.S.P. (6), zinc carbonate (1.6), cupric carbonate (0.3), potassium iodate (0.01), sodium selenite (0.01), chromium potassium sulfate■H2O (0.55), sucrose (118.03).

^cAIN-76A vitamin mix #300050: contains (g/kg mix): Thiamin■HCl (0.6), riboflavin (0.6), pyridoxine■HCl (0.7) niacin (3.0), calcium pantothenate (1.6), folic acid (0.2), biotin (0.02), vitamin B12 (0.1%; 1.0), vitamin A palmitate (500,000 iu/g; 0.8), vitamin D3 (400,000 iu/g (0.25), vitamin E acetate (500 iu/g; 10.0), menadione sodium bisulfite (0.08), sucrose (981.15).

Table 3.

Ingredients of the diets used in Experiment 3.

	Diet name abbreviation and Dyets catalog no.
Ingredient, kcal/g	C75-F5 Sucrose 104916	C75-F5 Starch 104917	C40-F40 Sucrose 104920	C40-F40 Starch 104921	C5-F75 Sucrose 104918	C5-F75 Starch 104919
Casein, 3.58	207.5	207.5	207.5	207.5	207.5	207.5
DL-Methionine, 4.0	3.0	3.0	3.0	3.0	3.0	3.0
Cornstarcha, 3.6	107.2	365.3	56.8	193.7	6.5	22.1
Dyetroseb, 3.8	33.8	115.3	17.9	61.2	2.0	7.0
Sucrose, 4.0	527.7	217.0	273.1	108.8	19.3	0.5
Vegetable shorteningc, 9.0	21.0	21.0	160.5	160.5	300.0	300.0
Cellulose, 0.0	52.4	23.5	233.8	217.9	414.2	412.4
Mineral mixd, 0.47	35.0	35.0	35.0	35.0	35.0	35.0
Vitamin mixd, 3.92	10.0	10.0	10.0	10.0	10.0	10.0
Choline chloride, 0.0	2.0	2.0	2.0	2.0	2.0	2.0
Dye, 0.0 (green/blue/red)	0.5	0.5	0.5	0.5	0.5	0.5
Total energy, kcal/kg	3.62	3.62	3.62	3.62	3.62	3.62
Percent energy as:
Protein %	20	20	20	20	20	20
Carbohydrate %	75	75	5	5	40	40
Fat %	5	5	75	75	40	40

Notes: Values in the body of the table are grams per kilogram diet. Diets were prepared by Dyets Inc. (Bethlehem, PA) in 50-kg lots.

^aSucrose is 81% of total carbohydrate in the high-sucrose diets and 35% of total carbohydrate in the high-starch diets.

^bDyetrose is a proprietary form of maltodextrin that facilitates pelleting the diet. The ratio of Dyetrose:cornstarch is 31.5% for all six diets.

^cPrimex brand

^dThe ingredients of AIN-76A mineral mix #200000 and vitamin mix #300050 are listed in Table 1

The diets were prepared and pelleted by Dyets Inc (Bethlehem, PA). They were shipped in 10-kg boxes to our facility, where they were kept in a cold room at 4 C. When needed, batches of 1-2 kg diet were transferred to plastic buckets with lids and kept in the vivarium at room temperature. Food hoppers were refilled weekly, before food intake tests (see below).

2.1.3. General Procedures

2.1.3.1. Body weight.

Beginning one week before the introduction of experimental diets, mice were weighed weekly. To do this, each mouse was removed from its home cage and placed in a plastic tub on a tared top-loading balance (with ± 0.1 g precision). It was then either returned to a new, clean cage or had its body composition assessed.

2.1.3.2. Body composition.

Body compositions were assessed on the day prior to the introduction of experimental diets and then every 4 weeks. To do this, each mouse was inserted into a Bruker Minispec LF110, which uses magnetic resonance technology to assess lean tissue weight, fat tissue weight, and water content [5]. Each mouse was placed into a plastic restraining tube and inserted into the core of the Minispec for ^~90 sec while body composition was assessed. The Minispec does not allow separate determinations of muscle and lean weight, or interstitial fat and individual fat depots.

2.1.3.3. Food intake.

Food intakes were assessed beginning on the day after the weekly body weight and/or composition measurement. Food given to each mouse was weighed then, after 24 h, reweighed. During these tests, the aspen chip bedding material scattered on the floor of each mouse’s cage was replaced with a corrugated cardboard sheet. The cardboard sheet allowed easy detection and collection of any spilled food, which was accounted for when calculating food intakes. In Experiment 1, some mice received a choice between two diets, which could be distinguished by color (Table 1) allowing spillage of each diet to be assessed separately. Also in Experiment 1, food intakes of the SWITCH group were measured on two consecutive days to capture intakes of both the C75-F5 diet and the C5-F75 diet.

2.1.4. Experiment Design

2.1.4.1. Experiment 1: Initial demonstration and CHOICE and SWITCH groups.

This experiment involved 96 male mice. All the mice were fed CHOW when they arrived at age 8-week old. At 11-week old, as part of another experiment, each mouse was videotaped in a clean cage while it was exposed to an innocuous odor (e.g., vanilla) for 3 min. When 14-week old, at the start of this experiment, the mice were assigned to one of six groups matched for body weight and body fat content. The groups differed in the food they received: (a) CHOW (n = 12); this group continued to eat Teklad 8604 diet, (b) C75-F5 (n = 16); this group ate a high-carbohydrate, low-fat diet described in Table 1, (c) C5-F75 (n = 16); this group ate a low-carbohydrate, high-fat diet that had the same energy density as the C75-F5 diet (Table 1), (d) C40-F40 (n = 16); this group ate a diet with equal amounts of carbohydrate and fat (by energy content), (e) CHOICE (n=20); this group had continuous access to both the C75-F5 diet and C5-F75 diet, and (f) SWITCH (n=16); this group had the C75-F5 and C5-F75 diets switched every 24 h; the C75-F5 diet was available on odd days and the C5-F75 diet was available on even days. Note that most groups contained 16 mice but the CHOICE group contained 20 mice. This was to increase statistical power because we anticipated the CHOICE group—the group with two sources of food—would be more variable than the other groups.

Body weights and food intakes were measured weekly and body compositions were assessed every 4 weeks for 16 weeks.

2.1.4.2. Experiment 2: Carbohydrate-fat dose-response in males and females.

This experiment involved 72 male and 72 female C57BL/6J mice. All 144 mice were fed CHOW when they arrived at age 9-week old. When 11-week old, the mice were assigned to nine diet groups, each comprising of 8 males and 8 females matched for body weight and body fat content within each sex. The diet groups differed in the food they received: One group of each sex continued to eat CHOW. The other 8 groups ate one of the 8 diets described in Table 2. Body weights and food intakes were measured weekly and body compositions were assessed at 0, 4 and 8 weeks.

Table 2.

Ingredients of the diets used in Experiment 2

	Diet name abbreviation and Dyets catalog no.
Ingredient, kcal/g	C75-F5 104586	C65-F15 104817	C55-F25 104818	C45-F35 104819	C35-F45 104820	C25-F55 104821	C-15-F65 104822	C5-F75 104587
Casein, 3.58	207.5	207.5	207.5	207.5	207.5	207.5	207.5	207.5
DL-Methionine, 4.0	3.0	3.0	3.0	3.0	3.0	3.0	3.0	3.0
Cornstarcha, 3.6	163.8	141.6	119.3	97.0	74.7	52.4	30.1	7.8
Sucrose, 4.0	546.1	471.9	397.6	323.3	249.0	174.4	100.3	26.1
Vegetable shorteningb, 9.0	21.0	62.9	104.8	146.8	188.7	230.6	272.5	314.5
Cellulose, 0.0	11.1	65.7	120.3	175.0	229.6	284.7	339.4	393.5
Mineral mixc, 0.47	35.0	35.0	35.0	35.0	35.0	35.0	35.0	35.0
Vitamin mixd, 3.92	10.0	10.0	10.0	10.0	10.0	10.0	10.0	10.0
Choline chloride, 0.0	2.0	2.0	2.0	2.0	2.0	2.0	2.0	2.0
Dye, 0.0 (green/blue/red)	0.5	0.5	0.5	0.5	0.5	0.5	0.5	0.5
Total energy, kcal/kg	3.77	3.77	3.77	3.77	3.77	3.77	3.77	3.77
Percent energy as:
Protein %	20	20	20	20	20	20	20	20
Carbohydrate %	75	65	55	45	35	25	15	5
Fat %	5	15	25	35	45	55	65	75

Notes: Values in the body of the table are grams per kilogram diet. Diets were prepared by Dyets Inc. (Bethlehem, PA) in 50-kg lots. Batches of 1-2 kg were stored in a cold room at 4 C until needed.

^aCornstarch is a constant 23% of carbohydrate across all 8 diets.

^bPrimex brand

^cdThe ingredients of AIN-76A mineral mix #200000 and vitamin mix #300050 are listed in Table 1

2.1.4.3. Experiment 3 and 4: Type of carbohydrate and type of fat.

Experiment 3 involved manipulation of the type of carbohydrate and Experiment 4 involved manipulation of the type of fat. There were 84 male C57BL/6J mice in Experiment 3 and 156 in Experiment 4. All were fed CHOW when they arrived at age 9-week old. When aged 11-week old, the mice were assigned to groups matched for body weight and body fat content; the groups differed in the type and amount of carbohydrate or fat they were fed. In Experiment 3, there were a total of 7 groups of mice. One group was fed CHOW; one set of three groups received C75-F5, C40-F40 or C5-F75 diets with the carbohydrate provided mostly as sucrose (81% sucrose and 19% cornstarch). The remaining three groups received the same C75-F5, C40-F40 or C5-F75 diets but with the carbohydrate provided mostly as cornstarch (35% sucrose and 65% cornstarch; Table 3). In Experiment 4, the types of fat tested were vegetable shortening (the same as in the other experiments), corn oil, palm oil, and canola oil. Each of these was presented in the C75-F5, C40-F40 and C5-F75 formulations (Table 1). Thus, there were a total of 13 groups of mice, including the group fed CHOW. Body weights and food intakes were measured weekly and body compositions were assessed at 0, 4, 8 and 12 weeks.

2.1.5. Statistical analyses

Food intake measurements, collected in grams, were converted to kilocalories using a value of 3.00 kcal/g for CHOW (15), 3.77 kcal/g for the semisynthetic diets used in Experiment 1, 2, and 4, and 3.62 kcal/g for the semisynthetic diets used in Experiment 3. In Experiment 1, preference scores of the CHOICE and SWITCH groups for the C5-F75 diet were calculated as C5-F75 intake/(C5-F75 intake + C75-F5 intake) x 100. All data from all mice were included in analyses except for one mouse in the C5-F75 group of Experiment 1, which was excluded because it died from unknown causes during the second week of the experiment.

Differences in body weights, energy efficiency scores (defined as the ratio of energy intake to body weight gain over the duration of the experiment), and body compositions were assessed using mixed-design two-way analyses of variance with between-subject factors of Diet Group and Sex (Experiment 2), Carbohydrate Type (Experiment 3) or Fat Type (Experiment 4). All analyses included a within-subject factor of Week. Secondary analyses included planned comparisons of each of the groups fed semisynthetic diets with the group fed CHOW.

Post hoc Fisher LSD tests were used to identify the source of interactions and main effects when these were significant. All analyses used a criterion for significance of p < 0.05. Initial analyses of the body weight and composition results of each experiment revealed that changes occurring over weeks gave few insights beyond the fact that differences among groups in body weight and composition became progressively larger as the experiments progressed. Consequently, we present a complete analysis for Experiment 1 but for Experiments 2-4, we present analyses based on the final body weight and body composition values only.

Food intakes were assessed in the same manner as were body weights and body compositions. In all four experiments, initial analyses of food intakes revealed there were relatively stable differences between groups but there was week-to-week variability most likely caused by fluctuations in vivarium temperature and other environmental perturbations. Consequently, to simplify presentation and analysis, food intakes were reanalyzed based on the average of all the intake measurements collected. These analyses are presented here.

3.1. RESULTS

3.1.1. Experiment 1: Initial demonstration and CHOICE and SWITCH groups

3.1.1.1. Body weight.

Body weight was influenced by diet and the duration of exposure to the diet [Group x Week interaction, F(35,623) = 16.6, p < 0.0001; Fig. 1]. Relative to the CHOW group, the C75-F5, C5-F75, C40-F40, and CHOICE groups weighed significantly more from week 10, 14, 2 and 6, respectively. The C75-F5 and C5-F75 groups never differed in body weight. The C40-F40 group weighed significantly more than did the C75-F5, C5-F75, and CHOICE groups from week 2 (in each case), and the CHOICE group weighed significantly more than did the C75-F5 and C5-F75 groups from week 4 or 6, respectively. The SWITCH group never differed from the CHOW group in body weight, and it weighed significantly less than did the C75-F5, C5-F75, C40-F40, and CHOICE groups from week 6, 10, 2, and 4, respectively. At the end of the 16-week diet exposure period, the CHOW, C75-F5, C5-F75, and SWITCH groups all had similar body weights, which were significantly less than those of the CHOICE group. Mice in the CHOICE group weighed significantly less than did those of the C40-F40 group [F(5,88) = 16.6, p < 0.0001; Fig 2A].

Fig. 1.

Experiment 1: Mean ± SEM body weights of six groups of male C57BL/6J mice fed chow (3.0 kcal/g; n = 12), high-carbohydrate, low-fat diet (C75-F5; 3.8 kcal/g; n = 16), low-carbohydrate, high-fat diet (C5-F75; 3.8 kcal/g; n = 15), a 50:50 mixture of the C75-F5 and C5-F75 diets (i.e., C 40-F40; n = 16), or the C75-F5 and C5-F75 diets provided as a simultaneous choice (CHOICE; n = 20), or switched daily (SWITCH; n = 16).

Fig. 2.

Experiment 1: Body composition and energy efficiency of six groups of male C57BL/6J mice fed for 16 weeks with chow (3.0 kcal/g; n = 12), high-carbohydrate, low-fat diet (C75-F5; 3.8 kcal/g; n = 16), low-carbohydrate, high-fat diet (C5-F75; 3.8 kcal/g; n = 15), an equal (i.e., 50:50) mixture of the C75-F5 and C5-F75 diets combined (C40-F40; n = 16), or the C75-F5 and C5-F75 diets provided as a simultaneous choice (CHOICE; n = 20), or switched daily (SWITCH; n = 16). Values are mean ± SEs. A. body weight, B. energy efficiency (energy intake/body weight gain), C. lean tissue weight, D. lean tissue weight as a proportion of body weight, E. fat tissue weight, and F. fat tissue weight as a proportion of body weight. Bars with different letters differed significantly according to Fisher LSD post hoc tests.

3.1.1.2. Body composition.

There were no group differences in body composition measured during the baseline period. At 4, 8, 12 and 16 weeks, group differences were stable, so we present only the results obtained at 16 weeks [lean tissue weight, F(5,88) = 10.4, p < 0.0001; lean tissue as a proportion of body weight, F(5,88) = 29.9, p < 0.0001; fat tissue weight, F(5,88) = 28.4, p < 0.0001; fat tissue as a proportion of body weight, F(5,88) = 29.9, p < 0.0001]. Differences among the groups in lean tissue weight and fat tissue weight are depicted in Figure 2C–E.

3.1.1.3. Food intake.

The C5-F75 and CHOICE groups consumed significantly more energy than did the CHOW and SWITCH groups, and significantly less energy than did the C75-F5 and C40-F40 groups [Fig. 3; F(5,88) = 18.2, p < 0.0001], The CHOICE group ate 54 ± 3% of its energy as C5-F75 diet (and thus 46 ± 3% as C75-F5 diet). Preferences of the CHOICE group for C5-F75 diet ranged from 29 – 82%. The correlation between C5-F75 diet preference and body weight gain was not significant, r = 0.04, n = 20. The SWITCH group ate exactly the same amount of food on days when fed C75-F5 diet (11.3 ± 0.2 kcal/day) versus days when they were fed C5-F75 diet (11.3 ± 0.2 kcal/day). Preference scores of the SWITCH group for the C75-F5 diet ranged from 48 – 53% and were not associated with body weight gain, r = 0.37, n = 16, p = 0.16 (NS).

Fig. 3.

Experiment 1: Mean ± SEs daily food intake of six groups of male C57BL/6J mice fed for 16 weeks with chow (3.0 kcal/g; n = 12), high-carbohydrate, low-fat diet (C75-F5; 3.8 kcal/g; n = 16), low-carbohydrate, high-fat diet (C5-F75; 3.8 kcal/g; n = 15), a, 50:50 mixture of the C75-F5 and C5-F75 diets (C40-F40; n = 16), or the C75-F5 and C5-F75 diets provided as a simultaneous choice (CHOICE; n = 20), or switched daily (SWITCH; n = 16). Bars with different letters differed significantly according to Fisher LSD post hoc tests. For the CHOICE and SWITCH groups, intake of C75-F5 diet is shown in green; intake of C5-F75 diet is shown in blue.

3.1.1.4. Energy efficiency.

Energy efficiency, the ratio of energy intake to body weight gain, differed significantly among the six groups, F(5,89) = 10.5, p < 0.0001. The C40-F40 and CHOICE groups had significantly lower efficiency scores than did the other four groups. The C75-F5 group had significantly lower energy efficient scores than did the C5-F75 group; the CHOW and SWITCH groups were intermediate between, and statistically indistinguishable from, the C75-F5 and C5-F75 groups (Fig. 2B).

3.1.2. Experiment 2: Carbohydrate-fat dose-response in males and females

3.1.2.1. Body weight.

At the end of the experiment, after 8 weeks of access to the diets, there was a significant interaction between Sex and Diet influencing body weight, F(8,126) = 2.28, p = 0.0256. This interaction had complex causes but it could be attributed largely to males fed diets in the middle of the range (i.e., the C35-F45, C45-F35 and C55-F25 groups) weighing more than did males fed diets with more extreme carbohydrate and fat contents; females fed diets in the middle of the range also weighed more than did females fed diets at the extremes, but the differences between these female groups was not as pronounced as it was between the corresponding male groups (Fig. 4A).

Fig. 4.

Experiment 2: Body composition and energy efficiency of male (blue) and female (pink) mice fed one of eight isocaloric diets differing in carbohydrate (CHO) and fat content, or CHOW (horizontal lines with shading shows mean ± SE for groups fed chow). Values are mean ± SE (n = 8). A. body weight, B. energy efficiency (energy intake/body weight gain), C. lean tissue weight, D. lean tissue weight as a proportion of body weight, E. fat tissue weight, and F. fat tissue weight as a proportion of body weight. For each panel, bars with different letters differed significantly (based on post hoc tests of same-sex groups; differences from the CHOW-fed group are not shown).

For both sexes combined, mice fed diets in the middle of the range weighed significantly more than did mice fed diets with more extreme proportions of carbohydrate and fat, F(8,126) = 12.2, p < 0.0001: Mice of both sexes fed the C35-F45 and C45-F35 diets weighed significantly more than did mice of both sexes fed the C25-F55 and C55-F25 diets, which in turn weighed significantly more than did mice of both sexes fed the C5-F75, C15-F55 and C65-F15 diets.

3.2.1.2. Body composition.

Mice fed diets with similar amounts of carbohydrate and fat had significantly more lean tissue and significantly more fat tissue than did mice fed diets with more extreme proportions of carbohydrate and fat [lean tissue weight, F(8,126) = 9.04, p < 0.0001; lean weight as a proportion of body weight, F(8,126) = 15.9, p < 0.0001; fat tissue weight, F(8, 126) = 17.3, p < 0.0001; fat tissue weight as a proportion of body weight, F(8,126) = 15.9, p < 0.0001; Fig 4C–F]. The effects of diet on fat tissue but not lean tissue were more pronounced in males than females [Sex x Diet interaction, lean tissue weight, F(8,126) = 1.18, p = 0.32, Fig. 4C; lean tissue weight as a proportion of body weight, F(8,126) = 3.10, p = 0.0032 (Fig 4D); fat tissue weight, F(8,126) = 4.59, p < 0.0001; Fig. 4E; fat tissue weight as a proportion of body weight, F(8,126) = 3.09, p = 0.0032; Fig. 4F]. Overall, males had more lean and fat tissue than did females [lean weight, F(1,126) = 936.3, p < 0.0001; lean weight as a proportion of body weight, F(1,56) = 5.70, p = 0.0188; fat weight, F(1,126) = 50.9, p < 0.0001; fat weight as a proportion of body weight, F(1,56) = 3.09, p = 0.0032].

3.1.2.3. Food intake.

Food intakes were influenced by a significant interaction between Sex and Diet, F(8,126) = 2.63, p = 0.0106, with a complex statistical cause (Table 4). Overall, males ate significantly less food than did females, F(1,126) = 14.6, p = 0.0002; this was particularly pronounced for the groups fed the C15-F65, C35-F45 and C65-F15 diets. The analysis involving both sexes combined revealed a significant effect of diet, F(8,126) = 2.63, p = 0.0106. This was due to mice fed diets in the middle of the range (i.e., C35-F45 and C45-F35) eating significantly more than did mice fed diets with more extreme proportions of carbohydrate and fat.

Table 4.

Average daily energy intakes of male and female mice fed CHOW or a range of isocaloric diets differing in carbohydrate and fat content (Experiment 1)

Diet	Males	Females	Sexes Combined
CHOW	12.8 ± 0.3a	13.6 ± 0.4ab	13.2 ± 0.3ab
C75-F5	12.7 ± 0.2a	13.2 ± 0.4a	12.9 ± 0.2a
C65-F15	12.7 ± 0.3a	15.0 ± 0.6c*	13.8 ± 0.4bc
C55-F25	13.5 ± 0.2abc	13.6 ± 0.4ab	13.6 ± 0.2abc
C45-F35	14.1 ± 0.3bc	14.5 ± 0.2bc	14.3 ± 0.2c
C35-F45	13.5 ± 0.3abc	14.7 ± 0.3c*	14.1 ± 0.3c
C25-F55	13.2 ± 0.4ab	14.3 ± 0.4abc*	13.7 ± 0.3bc
C15-F65	13.5 ± 0.5abc	14.1 ± 0.3abc	13.8 ± 0.3bc
C5-F75	14.3 ± 0.5c	13.4 ± 0.3ab	13.9 ± 0.2bc

Values are means ± SEs (kcal/d; n=8 per sex). Means with different letter superscripts differ significantly from diet groups of the same sex.

^*significantly different from males fed same diet.

3.1.2.4. Energy Efficiency.

Energy efficiency was significantly lower in males than females, F(1,126) = 15.9, p = 0.0001. It was significantly influenced by diet, F(8,126) = 7.41, p < 0.0001. Mice of both sexes fed the C45-F35 or C35-F45 diets had lower energy efficiency scores than did all the other groups except for mice fed C55-F25 diet. Mice fed the CHOW, C75-F5 or C5-F75 diets had higher energy efficiency scores than did all the other diet groups. Mice fed the C55-F25, C25-F55, C65-F15 or C15-F65 diets were intermediate between these two clusters. The interaction between sex and diet was not significant, F(8,126) = 1.87, p = 0.07 (Fig 4B).

3.1.3. Experiment 3: Type of carbohydrate: high sucrose vs. high starch

3.1.3.1. Body weight.

At the end of the 12-week trial, mice fed the high-sucrose or high-starch versions of the C40-F40 diet were equally and significantly heavier than were those fed the corresponding versions of the C75-F5 or C5-F75 diets, F(2,66) = 49.0, p<0.0001. There were no significant differences in body weight between mice fed the C75-F5 diets and C5-F75 diets, no significant differences between mice fed any of the high-sucrose diets and those fed the corresponding high-starch diets, F(1,66) = 0.81, p = 0.37, and no interaction between carbohydrate-fat amount and carbohydrate type, F(2,66) = 1.54, p = 0.22. Relative to the CHOW-fed group, only the groups fed C40-F40 diets weighed significantly more; the groups fed the C75-F5 and C5-F5 diets did not differ in weight from the CHOW-fed group (Fig. 5 and 6A).

Fig. 5.

Experiment 3: Body weights of male mice fed isocaloric diets differing in carbohydrate type (mostly sucrose or mostly cornstarch) and carbohydrate-fat amount. Body weights of the CHOW-fed group are redrawn in each panel to provide a comparison (n = 12/group).

Fig. 6.

Experiment 3: Body composition and energy efficiency of male mice after 12 weeks fed isocaloric diets differing in carbohydrate-fat amount, with carbohydrate provided as mostly sucrose or mostly starch. A. body weight, B. energy efficiency (energy intake/body weight gain), C. lean tissue weight, D. lean tissue weight as a proportion of body weight, E. fat tissue weight, and F. fat tissue weight as a proportion of body weight. Values are mean ± SEs (n = 16). Dashed horizontal lines show mean value of CHOW-fed group.

3.1.3.2. Body composition.

Lean tissue weight and lean weight as a proportion of body weight were influenced by the amount of dietary carbohydrate-fat [F(2,66) = 26.0, p < 0.0001 and F(2,66) = 63.5, p<0.0001, respectively], but not the type of carbohydrate [F(1,66) = 1.37, p = 0.25 and F(1,66) = 0.46, p=0.50, respectively], nor the interaction between these two factors [F(2,66) = 0.96, p = 0.39, F(2,66) = 3.08, p = 0.0526, respectively; Fig. 6C, 6D]. Similarly, fat tissue weight and fat weight as a proportion of body weight were influenced by the amount of dietary carbohydrate-fat, [F(2,66) = 63.1, p < 0.0001 and F(2,66) = 63.5, p < 0.0001, respectively], but not by the type of carbohydrate [F(1,66) = 0.32, p = 0.57, and F(1,66) = 0.46, p = 0.50, respectively] nor the interaction of these two factors [F(2,66) = 2.51, p = 0.09, and F(2,66) = 3.08, p = 0.0526 respectively; Fig. 6E, 6F]. The groups fed the C40-F40 diets had significantly heavier lean tissue and fat tissue than did the groups fed the C75-F5 or C5-F75 diets, which did not differ from each other. Relative to the CHOW-fed group, there were no significant differences from the other groups in lean tissue weight; however, the high-sucrose C40-F40, high-starch C40-F40, and high-starch C75-F5 groups had significantly heavier fat tissue than did the CHOW-fed group (Fig. 6).

3.1.3.3. Food intake.

Energy intake was influenced by a significant interaction between the type of carbohydrate (i.e., mostly sucrose or mostly starch) and the amount of dietary carbohydrate-fat in the diet [F(2,66) = 3.20, p = 0.0473; Table 5]. Mice ate significantly more of the high-sucrose version of the C40-F40 diet than the high-sucrose versions of the C75-F5 or C5-F75 diets; they ate significantly more of the high-starch version of the C40-F40 diet than the high-starch version of the C5-F75 diet but not the high-starch version of the C75-F5 diet. Mice fed the high-sucrose version of the C75-F5 diet ate significantly more than did cohorts fed the high-starch version of the C75-F5 diet; there were no differences in intake of the high-sucrose and high-starch versions of the C40-F40 or C5-F5 diets. Three groups of mice had calorie intakes significantly less than did the group fed CHOW (the high-sucrose C75-F5, high-sucrose C5-F75, and high-starch C5-F75 groups); intakes of the remaining three groups fed semisynthetic diets did not differ from the CHOW-fed group.

Table 5.

Mean ± SE energy intakes (kcal/d) of mice fed diets differing in carbohydrate type and carbohydrate-fat amount (Experiment 3)

	Diet Carbohydrate (C) and Fat (F) Content, %
Carbohydrate type	C75-F5	C40-F40	C5-F75
High sucrose	13.1 ± 0.1a	14.3 ± 0.3b	12.6 ± 0.3a
High starch	13.7 ± 0.2b	13.9 ± 0.2b	12.7 ± 0.2a

Notes: Each value is the mean ± SE of 12 1-day food intake measurements made weekly over 12 weeks (n = 12 mice per group). The interaction between dietary carbohydrate-fat amount and carbohydrate type was significant, F(2,66) = 3.20, p = 0.0473. Means with the same letter superscript did not differ significantly from each other.

High sucrose = 81% of carbohydrate as sucrose and 19% as starch; High starch = 35% of carbohydrate sucrose and 65% as starch (including Dyetrose maltodextrin; see Table 3).

The group fed CHOW ate 13.8 ± 0.2 kcal/d.

3.1.3.4. Energy efficiency.

Energy efficiency scores depended on carbohydrate-fat amount, F(2,66) = 11.9, p < 0.0001, with the C40-F40 diets being significantly less efficient than the C75-F5 and C5-F75 diets, which were equally efficient (Fig. 6B). The type of carbohydrate consumed had no significant effect on energy efficiency, F(1,66) = 3.49, p = 0.0663, and carbohydrate type did not interact with carbohydrate-fat amount, F(2,66) = 0.99, p = 0.37.

3.1.4. Experiment 4: Type of fat: vegetable shortening, corn oil, palm oil and canola oil

3.1.4.1. Body weight.

Body weight was influenced by fat type, F(3,132) = 3.37, p = 0.0206, and carbohydrate-fat amount, F(2,132) = 75.3, p < 0.0001, but the two factors did not interact, F(6,132) = 1.22, p = 0.30. Overall, mice fed corn oil and palm oil were significantly heavier than were mice fed vegetable shortening; mice fed corn oil were also significantly heavier than were mice fed canola oil. Mice fed any of the four C40-F40 diets were markedly and significantly heavier than were mice fed the corresponding C75-F5 or C5-F75 diets or CHOW; mice fed C75-F5, C5-F75 diets and CHOW did not differ significantly from each other in body weight (Fig. 7 and 8A).

Fig. 7.

Experiment 4: Body weights of mice fed isocaloric diets differing in fat type and carbohydrate-fat amount. Body weights of the CHOW-fed group are redrawn in each panel to provide a comparison (n = 12/group).

Fig. 8.

Experiment 4: Body composition of male mice after 12 weeks fed isocaloric diets differing in carbohydrate-fat amount, with fat provided as vegetable shortening, corn oil, palm oil or canola oil. A. body weight, B. Energy efficiency (energy intake/body weight gain), C. lean tissue weight, D. lean tissue weight as a proportion of body weight, E. fat tissue weight, and F. fat tissue weight as a proportion of body weight. Values are mean ± SEs (n = 16). There were no interactions between carbohydrate-fat amount and fat type. Dashed horizontal lines show mean value of CHOW-fed group.

3.1.4.2. Body composition.

Lean tissue weight and lean weight as a proportion of body weight were influenced by the amount of carbohydrate-fat in the diet [F(2,132) = 37.2, p < 0.0001 and F(2,132) = 134.2, p < 0.0001, respectively] but not by the type of fat [F(3,132) = 2.54, p = 0.06 and F(3,132) = 1.34, p = 0.35, respectively] or their interaction [F(6,132) = 0.63, p = 0.70 and F(6,132) = 2.32, p = 0.62, respectively; Fig. 8B and 8D]. Irrespective of the type of fat eaten, mice fed the C40-F40 diets had significantly greater lean tissue weights than did mice fed the C5-F75 or C75-F5 diets. Fat tissue weight and fat weight as a proportion of body weight were influenced by the amount of fat eaten [F(2,132) = 114.4, p < 0.0001 and F(2,132) = 134.2, p < 0.0001, respectively] , and type of fat eaten [F(3,132) = 5.27, p = 0.0018 and F(3,132) = 6.65, p = 0.0003, respectively] but not by the interaction of these two factors [F(6,132) = 1.90, p = 0.08; F(6,132) = 2.32, p = 0.62, respectively; Fig. 8C and 8E]. The significant difference related to dietary fat type was due to mice fed corn oil or palm oil having heavier body fat than did mice fed vegetable shortening; other differences in fat weight due to eating different types of fat were not significant. Irrespective of the type of fat eaten, mice fed the C40-F40 diets had significantly greater fat weights than did mice fed the C5-F75 or C75-F5 diets, which did not differ from each other.

3.1.4.3. Food intake.

The interaction between fat type and carbohydrate-fat amount was not significant, F(6,132) = 1.39, p= 0.22 (Table 6). Overall, mice ate more C40-F40 and C5-F75 diet than C75-F5 diet, F(2,132) = 8,68, p = 0.0003. They also ate significantly more corn oil, palm oil, and canola oil than vegetable shortening, F(3,132) = 3.42, p = 0.0192.

Table 6.

Mean ± SE energy intakes (kcal/d) of mice fed diets differing in fat type and carbohydrate-fat amount (Experiment 4)

	Diet Carbohydrate (C) and Fat (F) Content, %
Fat type	C75-F5	C40-F40	C5-F75
Vegetable shortening	13.1 ± 0.2	13.7 ± 0.2	13.0 ± 0.2
Corn oil	13.0 ± 0.3	14.2 ± 0.4	14.6 ± 0.4
Palm oil	13.5 ± 0.2	14.4 ± 0.3	14.2 ± 0.5
Canola oil	13.1 ± 0.1	13.7 ± 0.3	13.7 ± 0.4

Each value is the mean of 12 1-day food intake tests conducted weekly over 12 weeks (n = 12 mice per group). The interaction between fat type and carbohydrate-fat amount was not significant, F(6,132) = 1.39, p = 0.224. Overall, the mice ate more C40-F40 and C5-F75 than C75-F5, F(2,132) = 8,68, p = 0.0003. They also ate more corn oil, palm oil, and canola oil than vegetable shortening, F(3,132) = 3.42, p = 0.0192. The group fed CHOW ate 12.9 ± 0.2 kcal/d.

3.1.4.4. Energy efficiency.

Energy efficiency depended on carbohydrate-fat amount, F(2,132) = 19.1, p < 0.0001, with the C40-F40 diets yielding significantly lower efficiency scores than did the C75-F5 and C5-F75 diets, which were equally efficient. The type of fat consumed had a marginally significant effect on energy efficiency scores, F(3,132) = 2.75, p = 0.0452, an effect due to the groups fed canola oil having higher scores than did the groups fed corn oil or palm oil. There was no significant interaction between carbohydrate-fat amount and fat type, F(6,132) = 1.55, p = 0.83.

4.1. DISCUSSION

These experiments provide several rigorous demonstrations that when dietary energy density is held constant, body weight and body fat accumulation depend on the proportion of carbohydrate and fat consumed. Mice gained weight similarly whether they were fed a C75-F5 diet or an isocaloric C5-F75 diet. This is consistent with the source of energy being unimportant for body weight gain. However, mice fed C40-F40 diet gained substantially more weight than did those fed either of the isocaloric C75-F5 or C5-F75 diets. The higher body weight of the mice fed C40-F40 diet was a function of the proportion of carbohydrate-to-fat but, at least for the examples we tested, not the type of carbohydrate (high-sucrose or high-starch) or fat (vegetable shortening, corn oil, palm oil, canola oil). Similarly, mice allowed to mix their own combination of carbohydrate and fat by choosing between the C75-F5 and C5-F75 diets selected ^~50% from each source and gained almost as much weight as did mice fed the C40-F40 diet. Mice fed the C75-F5 and C5-F75 diets on different days gained no more weight than did the groups fed C75-F5 diet alone or C5-F75 diet alone. The implication is that obesity is exacerbated by consumption of a 50:50 combination of carbohydrate and fat, providing these are consumed in temporal proximity.

Critically, weight gain depended on the combined availability of carbohydrate and fat, independent of the diets’ protein, micronutrient or energy content because in all four experiments protein, micronutrients and energy content of the diets remained constant. Thus, our results contradict the assertion of Hu et al (24) that dietary fat, but not protein or carbohydrate, regulates energy intake and causes adiposity in mice; these investigators did not dissociate dietary fat from its energy content. Our results also cannot be explained as due to leverage by protein imbalance [i.e., a response to “nutritional geometry” (43–46)] because the diets produced differences in body weight despite having identical protein content.

Ramirez and Friedman (37) demonstrated that the combination of high-carbohydrate, high-fat and high-energy content is particularly conducive to obesity in rodents. In their study, rats fed a “mixed, low-energy” diet similar to the C40-F40 diet used here had a nonsignificant tendency to gain more weight and carry more adipose tissue than did rats fed isocaloric high-carbohydrate, low-fat or low-carbohydrate, high-fat diets. This discrepancy (i.e., a nonsignificant trend versus statistical significance) could be a species difference but we suspect it is more likely due to the lower statistical power of the study by Ramirez and Friedman [i.e., groups of 8-9 outbred female rats tested for 5 weeks (37) versus groups of 12 or more inbred mice tested for 8-16 weeks here].

We explicitly manipulated dietary carbohydrate and fat content but we are well-aware this also involved manipulation of the quantity of cellulose added to the high-fat diets to lower their energy density. Cellulose does not affect the body weight of rodents unless dietary energy content is diluted to a level that bulk capacity of the gut becomes a problem (1,34,35,37), which was not an issue here. A further argument against the contribution of cellulose to body weight gain is that the highest weight gains occurred in the middle of the range we tested; the C75-F5 diet, which contained very little cellulose, and the C5-F75 diet, which contained lots of cellulose, both caused less weight gain than did the C40-F40 diet, which contained an intermediate amount of cellulose. Thus, any influence of cellulose is not monotonically “dose-related.” Our manipulation of carbohydrate and fat content also altered the diets’ orosensory characteristics, but we doubt this is important because diet palatability does not influence body weight (33,54). Moreover, here we can rule out a contribution of orosensory characteristics to body weight because mice given a choice between the C75-F5 and C5-F75 diets became fatter than those given the same foods alone. It is implausible that the palatability of each diet is influenced by simply having the other diet available, particularly when neither diet was more-preferred in 24-h two-cup choice tests.

Our finding that mice fed a mixture of the C75-F5 and C5-F75 diets were fatter than those fed either diet alone would be less interesting if the C75-F5 or C5-F75 diets had too little or too much of a macronutrient to support normal growth. However, this was not the case. First, mice fed the C75-F5 and C5-F75 diets had body weights that fell naturally at the tails of a smooth inverted U-shaped function relating diet carbohydrate-fat to body weight (Fig. 4). We would expect a more abrupt effect on body weight, particularly lean tissue weight, if the diets at the extremes of this distribution were pathologic, for example, if the C5-F75 diet was ketogenic. Second, mice fed the C75-F5 diet or the C5-F75 diet gained at least as much or more weight than did CHOW-fed control mice. The finding that the C75-F5 diet and C5-F75 diet produced equivalent weight gains in C57BL/6J mice has also been observed with (NZB/NZW)F1 mice fed slightly different diet formulations (27). Our use here of a mixture of sucrose and cornstarch as the source of carbohydrate is unlikely to be critical because similar weight gains were obtained whether the carbohydrate was largely sucrose or largely cornstarch in C57BL/6J (and A/J) mice [Experiment 3 and (48)].

In one experiment, we compared mice fed palm oil, canola oil, corn oil and vegetable shortening, which we chose to exemplify saturated, monounsaturated, polyunsaturated, and trans fats, respectively. The type of fat eaten had small but significant effects on body weight and fat stores; mice fed corn oil or palm oil gained more weight and fat than did mice fed vegetable shortening. However, fat type did not influence the body weight gain produced by the C40-F40 diet relative to either the C75-F5 or C5-F75 diets. We thus conclude that the type of fat has little impact on body weight and body composition relative to the amount of fat in the diet.

Generally, there was a strong correspondence between body weight assessed using a top-loading balance and fat weight assessed by magnetic resonance with a Bruker Minispec LF110. In groups that gained the most body weight there were also gains in lean tissue weight—mice fed the C40-F40 diet had heavier lean tissue as well as fat tissue—but the changes in lean tissue were tiny. The sum of lean tissue weight plus fat tissue weight estimated by the Minispec was sometimes greater than body weight obtained by measurement with a top-loading balance. We suspect that these discrepancies were due to calibration errors affecting extremely fat mice, although there may also be errors caused by differences in body fluid content (21). The Minispec is designed for body composition analysis of mice and several studies have used it [review (5)] but we are unaware of a direct validation study. We note that, whatever the cause of the discrepancy between Bruker Minispec and top-loading balance, the effects of the diets we observed here dwarf any errors in body composition assessment.

One finding that remains to be explained is the relationship between body weight gain and food intake. On the one hand, mice that gained the most weight (e.g., the C40-F40 groups in Experiment 1, 3 and 4 and the CHOICE group in Experiment 1) ate the most food, and mice that gained the least weight (e.g., the C75-F5 and C5-F75 groups in Experiment 1, 3 and 4 and SWITCH group in Experiment 1) ate the least food. Thus, there appears to be a relationship between hyperphagia and obesity. On the other hand, in some cases (i.e., in Experiment 1 and 2) mice fed the C75-F5 diet ate significantly more than did mice fed the C5-F75 diet despite having statistically identical body weights. This is consistent with the lower efficiency of conversion of dietary carbohydrate than fat to body weight described in rats [e.g., (13)], perhaps reflecting higher thermogenesis and thus greater energy needs in response to carbohydrate than fat [(58) but see (40)]. Analyses here of “energy efficiency” (defined here as energy intake/body weight gain) provided little insight; groups of mice that gained the most weight had the lowest energy efficiencies. It will require additional studies measuring diet effects on energy loss to better understand the energy balance of our mice.

Another finding that remains to be understood is that in the carbohydrate-fat “dose-response’ experiment, the peak of weight gain occurred in mice fed approximately 50:50 mixtures of carbohydrate and fat (Fig. 4). Why should these proportions be more effective than any other combination of carbohydrate and fat? The coincidence that the 50:50 carbohydrate:fat ratio was most effective at causing obesity raises the possibility that the underlying physiological mechanism is based on energy stoichiometry rather than, for example, carbohydrate-induced insulin spikes, for which there is no need to postulate the participation of equal amounts of carbohydrate and fat to convey maximal weight gain.

One impetus for this study was to investigate why cafeteria diets are particularly effective at stimulating obesity [e.g., (41)]. At least part of this phenomenon involves the “variety effect”—the idea that total food intake is greater if two or more foods are presented than if just one food is presented (38,39). This has been studied frequently in humans but only occasionally in laboratory animals. It is usually explained as due to attenuation of sensory specific satiety although evidence argues against this; providing rats with a variety of flavored foods with identical nutrient compositions does not affect body weight gain (33). The goal here was to investigate whether providing foods with different nutrient compositions but identical energy densities affected weight gain. The results of the CHOICE group indicate that this is, in fact, the case. Mice that could choose from C75-F5 and C5-F75 diets gained significantly more weight than did mice fed either diet alone. Indeed, they gained almost as much body weight and body fat as did mice fed the composite C40-F40 mixture. Interestingly, the CHOICE group selected almost exactly 50% of its food as C75-F5 diet and 50% as C5-F75 diet. This is the result expected if the mice were indifferent to the two diets, although it is straining credulity to argue that they could not distinguish between the diets based on their sensory properties. Instead, we think it likely that the mice could distinguish between the two diets but considered them to be of equal value. This is predicted by the hypothesis that food preference is determined by fuel oxidation (51,55) but, whatever the explanation, the influence of variety on body weight can be explained by the mouse’s choice of a balanced mixture of carbohydrate- and fat-containing foods rather than, or in addition to, enhanced sensory stimulation.

It is informative to compare the results of the CHOICE and SWITCH groups in Experiment 1. These two groups both had the opportunity to eat C75-F5 and C5-F75 diet yet the CHOICE group ate significantly more food and accumulated twice the body fat than did the SWITCH group. The operational difference between the groups was that for the CHOICE group, C75-F5 and C5-F75 diet access was simultaneous; for the SWITCH group, it was successive. However, the CHOICE group can be considered a SWITCH group, in the sense that each mouse repeatedly selects first one, then the other diet, switching between the two (in a pattern we did not attempt to quantify). We do not know whether this switching occurs between every meal or even within each meal [see (38)]. Nevertheless, the difference in outcome between the CHOICE and SWITCH groups implies that the interaction between carbohydrate and fat influencing body weight must be a rapid and transient process. We believe future attention should be paid to the rapid switching between carbohydrate and fat fuels as a contributor to obesity.

The SWITCH group, which alternated between the C75-F5 and C5-F75 diets every 24 h, ate significantly less food and gained significantly less body weight than did the groups fed C75-F5 alone or C5-F75 alone. There must be something about switching between high-carbohydrate and high-fat diets that inhibits food intake and weight gain. Several studies have shown that time-restricted feeding prevents obesity even without reductions of energy intake (8,9,22) but to our knowledge, this is the first demonstration that restricting macronutrient availability has the same effect. The microbiomic and physiological accommodations that occur with prolonged exposure to a single diet may be prevented by daily diet switching. One possibility is that meal-related changes in the gut microbiome counter adaptations to the high-carbohydrate and high-fat diets (7,11). Another is that the pathways for fuel oxidation become conditioned to accommodate the change in fuel availability. We also cannot rule out the mundane possibility that the act of changing the food of these mice every 24 h disturbs eating patterns in a manner that leads to lower body weight gain. To account for this albeit unlikely possibility, future studies should include a control group subjected to the same disturbances but with the same type of food returned every day.

Our decision to test C57BL/6J mice was based on the ubiquity of this strain in genetic and obesity research [e.g., (10,12,23,24,28,32,36,46,49,54,56)]. Compared with other strains, the C57BL/6J strain gains weight robustly when fed high-energy, high-fat diets (19,29,53,57). Somewhat surprisingly given that the C57BL/6J strain is inbred and so individuals are “genetically identical,” this strain shows a heterogeneous weight gain response to dietary fat (4,6). However, this variability did not prove to be a concern for the present work; there was low inter-subject variability in the body weights of mice fed the same diet. Our decision to test only males in Experiment 1 was primarily to save money; males are less expensive to purchase. There are also no worries about the estrus cycle increasing variability, so it was prudent to confine testing to males. The results of Experiment 1 were sufficient for us to obtain funding to cover the cost of testing both sexes in Experiment 2. This experiment showed that males and females responded similarly to a range of diets differing in carbohydrate and fat, albeit with females showing a muted response relative to males, so there was no justification for testing both sexes in Experiments 3 and 4. We note that it is a strength of our paper to include any results from both sexes; most similar work confines testing to only one sex. Nevertheless, the results reported here may not generalize to strains, species and sexes we did not test.

An obvious question is whether these results have implications for humans. Our finding that C40-F40 diet, which approximates the composition of the Western diet, causes greater obesity than either C75-F5 diet (which is akin to a Pritikin diet) or C5-F75 diet (which is akin to an Atkins diet) argues that both the Pritikin and the Atkins diets would be more beneficial for keeping off the weight than the typical Western diet; neither has the advantage, but choosing and sticking to one or the other exclusively is important. However, the results of our SWITCH group, which gained less body weight than did mice fed other combinations of the C75-F5 and C5-F75 diets, argues there may be advantages to a weight-loss diet regimen that involves switching daily between high-carbohydrate and high-fat sources. To our knowledge, such a regimen has not been popularized.

Finally, we stress that our experiment was designed to eliminate food energy and protein content as factors in the regulation of body weight, but this is not to deny that these contribute to body weight gain under normal circumstances [e.g., (14,16,26,56)]. In our view, the most effective diets to cause obesity have the combination of low-protein, high-carbohydrate, high-fat, and high-energy. The findings made here demonstrate that carbohydrate and fat act by virtue of more than simply their energy content; there must be a metabolic interaction. A prime mechanism for this is carbohydrate-induced hyperinsulinemia facilitating adipose tissue storage of fat fuels [see (17,18,20,31,50)] but it is probably simplistic to believe this is the only mechanism [see (23,47)]. It will take concerted experimentation to identify and characterize the underlying physiological and neural mechanisms. It is surprising that such a weighty issue remains to be investigated.

HIGHLIGHTS.

• Mice fed 50/50 carbohydrate:fat are fatter than those fed other proportions

• Mice choose a 50/50 carbohydrate:fat diet and become obese

• The type of carbohydrate or fat has little influence on body weight